Lei Wang

Lei Wang, Benedict Yeoh, Jun Wah Ng

Synthetic voice and splicing audio clips have been generated to spoof Internet users and artificial intelligence (AI) technologies such as voice authentication. Existing research work treats spoofing countermeasures as a binary classification problem: bonafide vs. spoof. This paper extends the existing Res2Net by involving the recent Conformer block to further exploit the local patterns on acoustic features. Experimental results on ASVspoof 2019 database show that the proposed SE-Res2Net-Conformer architecture is able to improve the spoofing countermeasures performance for the logical access scenario. In addition, this paper also proposes to re-formulate the existing audio splicing detection problem. Instead of identifying the complete splicing segments, it is more useful to detect the boundaries of the spliced segments. Moreover, a deep learning approach can be used to solve the problem, which is different from the previous signal processing techniques.

Lei Wang, Rong Tong, Cheung Chi Leung et al.

This paper provides an overall introduction of our Automatic Speech Recognition (ASR) systems for Southeast Asian languages. As not much existing work has been carried out on such regional languages, a few difficulties should be addressed before building the systems: limitation on speech and text resources, lack of linguistic knowledge, etc. This work takes Bahasa Indonesia and Thai as examples to illustrate the strategies of collecting various resources required for building ASR systems.

Lei Wang, Rong Tong

One of the challenges in automatic speech recognition is foreign words recognition. It is observed that a speaker's pronunciation of a foreign word is influenced by his native language knowledge, and such phenomenon is known as the effect of language transfer. This paper focuses on examining the phonetic effect of language transfer in automatic speech recognition. A set of lexical rules is proposed to convert an English word into Mandarin phonetic representation. In this way, a Mandarin lexicon can be augmented by including English words. Hence, the Mandarin ASR system becomes capable to recognize English words without retraining or re-estimation of the acoustic model parameters. Using the lexicon that derived from the proposed rules, the ASR performance of Mandarin English mixed speech is improved without harming the accuracy of Mandarin only speech. The proposed lexical rules are generalized and they can be directly applied to unseen English words.

Kuiye Ding, Fanda Fan, Zheya Wang et al.

In the World Wide Web, reliable time series forecasts provide the forward-looking signals that drive resource planning, cache placement, and anomaly response, enabling platforms to operate efficiently as user behavior and content distributions evolve. Compared with other domains, time series forecasting for Web applications requires much faster responsiveness to support real-time decision making. We present KAIROS, a non-autoregressive time series forecasting framework that directly models segment-level multi-peak distributions. Unlike autoregressive approaches, KAIROS avoids error accumulation and achieves just-in-time inference, while improving over existing non-autoregressive models that collapse to over-smoothed predictions. Trained on the large-scale corpus, KAIROS demonstrates strong zero-shot generalization on six widely used benchmarks, delivering forecasting performance comparable to state-of-the-art foundation models with similar scale, at a fraction of their inference cost. Beyond empirical results, KAIROS highlights the importance of non-autoregressive design as a scalable paradigm for foundation models in time series.

Xin Feng, Lei Wang

In this paper, we present how to hybridize a Word2vec model and an attention-based end-to-end speech recognition model. We build a phoneme recognition system based on Listen, Attend and Spell model. And the phoneme recognition model uses a word2vec model to initialize the embedding matrix for the improvement of the performance, which can increase the distance among the phoneme vectors. At the same time, in order to solve the problem of overfitting in the 61 phoneme recognition model on TIMIT dataset, we propose a new training method. A 61-39 phoneme mapping comparison table is used to inverse map the phonemes of the dataset to generate more 61 phoneme training data. At the end of training, replace the dataset with a standard dataset for corrective training. Our model can achieve the best result under the TIMIT dataset which is 16.5% PER (Phoneme Error Rate).